What is prediction accuracy?

Prediction accuracy is a concept in personal epistemology: The ability to see clearly — not optimistically, not pessimistically, but accurately — is rarer and more valuable than most technical skills. Calibrated perception compounds into better decisions, and better decisions compound into better outcomes at every timescale.

Example: Two product managers evaluate a new feature launch. One is uncalibrated: she estimates 80% confidence that the feature will hit adoption targets, ignores base rates for similar launches (L-0151), doesn't run a pre-mortem (L-0153), and dismisses early negative signals as noise. The feature misses targets by 40%, and she is genuinely surprised. The other is calibrated: he estimates 55% confidence, factors in the base rate that only 30% of comparable features hit initial targets, runs a pre-mortem that identifies three likely failure modes, and builds measurement into the first sprint. The feature also underperforms — but he planned for that contingency, pivots in week two, and recovers half the gap. Same company. Same feature. Same data. The calibrated perceiver didn't predict the future. He perceived the present accurately enough to prepare for multiple futures.

This concept is part of Phase 8 (Perceptual Calibration) in the How to Think curriculum, which builds the epistemic infrastructure for perceptual calibration.